Mechanics
http://ijd.sagepub.com/
Published by:
http://www.sagepublications.com
Additional services and information for International Journal of Damage Mechanics can be found at:
Subscriptions: http://ijd.sagepub.com/subscriptions
Reprints: http://www.sagepub.com/journalsReprints.nav
Permissions: http://www.sagepub.com/journalsPermissions.nav
Citations: http://ijd.sagepub.com/content/22/7/967.refs.html
What is This?
interface
R Li and LZ Sun
Abstract
A viscoelastic interface model is proposed to study the effective mechanical responses of nanocomposites
filled with carbon nanotubes. It is incorporated with the Mori–Tanaka method and the elastic-viscoelastic
correspondence principle. The effective dynamic stiffness and damping of randomly oriented nanocom-
posites are investigated with the consideration of imperfect interfacial condition between the filler and
matrix as well as the concentration and aspect ratio of fillers. Comparisons are performed between the
model-based simulation and the experimental data for a specific type of multi-walled carbon nanotube
reinforced elastomer nanocomposites to demonstrate the potential of the proposed framework.
Keywords
Nanocomposites, carbon nanotubes, micromechanics, viscoelasticity, imperfect interface, dynamic mech-
anical analysis
Introduction
Interfacial bonding condition between the inhomogeneities and matrix plays an important role in
determining the overall mechanical properties of composite materials. Imperfect interfaces may arise
for various reasons such as thin coating layers, interfacial debonding, surface chemical reactions, etc.
Mathematically speaking, for a perfect interface the displacement and traction fields are both con-
tinuous across the boundary of two phases, while for an imperfect interface this is not the case.
Generally the modeling efforts which account for the effect of imperfect interfaces can be divided
into two major categories. The first category of models assumes interfaces where the displacement
or traction field is discontinuous with no thickness, and is usually referred to as interface models.
Department of Civil and Environmental Engineering, University of California, Irvine, CA, USA
Corresponding author:
LZ Sun, Department of Civil and Environmental Engineering, University of California, Irvine, CA 92697-2175, USA.
Email: lsun@uci.edu
The other category of models describes the interface as a layer of a certain thickness whose proper-
ties are different from those of either matrix or inhomogeneities, and is usually referred to as inter-
phase models (Hashin, 2002). For the above models to be successful, appropriately assigning the
interface/interphase parameters is the key factor, and commonly a difficult task which requires
precise experimental characterization.
The studies involving interface models originate from the work by Benveniste (1985), in which the
mathematical framework of linear spring interfaces has been formulated. Aboudi (1987) has applied
finite element unit cell method to investigate the interfacial effect for particulate and unidirectional
long-fiber composites. Qu (1993a, 1993b) has derived a modified Eshelby tensor for the case of
imperfect interfaces and incorporated that into the application of Mori-Tanaka method. Recently, a
three-stage nonlinear interfacial cohesive law has been proposed by Tan et al. (2008). Other works
can be seen, for example Achenbach and Zhu (1989), Hashin (1991a), Jasuik et al. (1987), and
Pagano and Tandon (1990). It should be noted that in many linear spring type of interface
models, only sliding along tangential direction is allowed, while relative motion along normal dir-
ection is prohibited to avoid material interpenetration and simplify numerical manipulations. The
summary of other linear interface models including free sliding model, dislocation-like model and
the interface stress model can be seen from, for example, Duan et al. (2005).
Interphase models have also been extensively studied and applied to fiber and particle reinforced
composites. Representative works include the studies on elastic interphases (Achenbach and Zhu,
1990) and those on viscoelastic interphases (Fisher and Brinson, 2001; Hashin, 1991b; Li and Weng,
1996), in which micromechanics, finite element method, and boundary element method are utilized.
The interface and interphase models are not completely independent of each other; in fact, Hashin
(1990, 1991b) has proved that the spring constant parameters in interface models can be calculated
in terms of interphase thickness and elastic properties in interphase models.
In addition to interface and interphase models, Duan et al. (2007) have proposed a unified scheme
to replace either interphases or interfaces by equivalent homogeneous particles or fibers based on
energy equivalency. An evolutionary damage framework which replaces debonded isotropic par-
ticles by perfectly bonded equivalent orthotropic ones has been proposed by Liu et al. (2004), which
was later applied to functionally graded materials by Paulino et al. (2006). Further investigations of
micromechanics-based interfacial debonding and damage of composites have been studied by Jain
and Ghosh (2009), Ju and Ko (2008), Ju and Yanase (2009, 2011), Ju et al. (2006, 2008, 2009), Lee
and Pyo (2008), Liu et al. (2006), Kim and Lee (2011), Sun et al. (2003a, 2003b), and Yanase and Ju
(2012), among others.
Linear elastic interface models have been extended to linear viscoelastic interface models by
replacing the interfacial linear spring with its viscoelastic counterpart, which is time and frequency
dependent. Hashin (1991c) has studied the creep and relaxation behavior of unidirectional fiber
composites with viscoelastic interfaces and analytical solutions have been given. Gosz et al. (1990)
and Matzenmiller and Gerlach (2004) have applied unit cell method to numerically investigate the
compliant fiber-matrix bond in fiber composites.
Carbon nanotube (CNT) reinforced nanocomposites have proved to demonstrate improved
mechanical performance including stiffness, strength, and damping when compared with neat
matrix materials (e.g. Schadler et al., 2007; Sun et al., 2009), due to the extraordinary properties
of CNTs including nanometer size, high aspect ratio, and high modulus. However, good dispersion
of CNTs and bonding between CNTs and matrix remain challenging tasks. CNT clustering and
poor bonding often lead to less efficient reinforcement in stiffness and strength (Xie et al., 2005).
Nevertheless, imperfect interfaces within nanocomposites allow for interfacial slippage and friction
where ij ðSþ Þ (or ui ðSþ Þ) and ij ðS Þ (or ui ðS Þ) are respectively the values of stress (or displacement)
approaching the interface S from outside and inside . ij is a second-rank tensor representing the
spring compliance of the interface, which is assumed symmetric and positive infinite for simplicity.
We can see that ij ¼ 0 stands for perfect interface and ij ! 1 completely debonded interface.It
can be written in the following form
where and represent the tangential and normal compliance of the interface, respectively. When
6¼ 0 and ¼ 0 are imposed, no separation between the two phases will occur but only tangential
sliding. This approximation cannot only prevent the unphysical scenario of material interpenetra-
tion, but also greatly simply the mathematical manipulation.
The concept of viscoelastic interfaces have first appeared in the works by Gosz et al. (1991) and
Hashin (1991c). The interfaces are no longer represented by linear springs, but their viscoelastic
counterparts, which can be phenomenologically visualized as a combination of linear springs and
dashpots. For a viscoelastic interface, the interfacial compliance is a function of time instead of a
constant. The second-rank tensor ij is now time dependent as
and consequently the interfacial displacement jump is time dependent as well, which can be written
in an integral form
Z t
@jk ð Þ
ui ðtÞ ¼ ij ðt Þ d nk : ð5Þ
1 @
In this article, the assumption of no interfacial separation in the normal direction, i.e. ðtÞ ¼ 0, is
adopted, and starting from here the tangential interfacial compliance will be substituted by inter-
facial compliance for conciseness.An illustration of this type of viscoelastic interface is given in
Figure 1, where the interfacial compliance is simulated by a standard linear solid model.
Composite materials show time-dependent modulus in creep and relaxation tests, and frequency-
dependent modulus and damping under dynamic loads, provided that one or more phases are
viscoelastic, or viscoelastic interfaces exist. Laplace transform has been used by Gosz et al.
(1991), Hashin (1991c), and Matzenmiller and Gerlach (2004) to apply the correspondence principle
to composites with linear viscoelastic interfaces. The problem is first solved in the Laplace domain
and the solution is inverted back into the time domain, therebythe time-dependent creep and relax-
ation behavior is investigated. It should be noted that the solution in the Laplace domain has in fact
no physical meaning. Although the dynamic modulus and damping can be obtained from thecon-
version of the time-dependent modulus, this process is nevertheless tedious and mathematically
complicated.
The other way to apply the correspondence principle is through Fourier transform (Christensen,
1980). The stress–strain relations of the linear viscoelastic phases are transformed into frequency
domain, and the solution in that domain can be obtained similar to the elastic solution. Since
viscoelastic materials do have frequency-dependent complex modulus, the frequency domain solu-
tion for modulus has direct physical meaning and does not need to be transformed back to time
domain. Therefore in case of finding harmonic viscoelastic solutions, e.g. effective complex modulus
of the composite, we can simply replace the elastic modulus of each phase by the corresponding
complex modulusin the elastic solutions. This process is simple, however, of great utility.
We now apply Fourier transform
Z 1
fð!Þ ¼ fðtÞei!t dt ð6Þ
1
Given that ðtÞ ¼ 0, and following the similar fashion of complex compliance of linear viscoelas-
tic materials we can write the complex interfacial compliance in the frequency domain as
where the storage (real) and loss (imaginary) parts are respectively
Z 1
0 ð!Þ ¼ ð1Þ ! ½ð1Þ ðt0 Þ sinð!t0 Þdt0 ð9Þ
0
Z 1
00 ð!Þ ¼ ! ½ð1Þ ðt0 Þ cosð!t0 Þdt0 ð10Þ
0
Here ð1Þ is the equilibrium compliance at infinite time. Then equation (7) can be written in the
following form
ui ð!Þ ¼ ð!Þ ij ni nj jk ð!Þnk ð11Þ
which possesses the equivalent form as equation (2) if letting ¼ 0. As a result,in addition to
replacing the elastic modulus of each phase by corresponding complex modulus, we may further
use ð!Þ to replace the elastic to get the complex modulus of composites with both linear visco-
elastic phases and interfaces, if the solution for the elastic case is available. Equation (8) can be also
written as
ð!Þ ¼ 0 ð!Þ 1 i½00 ð!Þ=0 ð!Þ ð12Þ
where an alternative form of the complex interfacial compliance is given. 00 ð!Þ=0 ð!Þ, or interfacial
loss factor, is comparable to the loss factor tan of viscoelastic phases and indicates the interfacial
damping capacity.
where it can be found that the perturbed strain field is also uniform in . For fiber-reinforced
composite, Sijkl would be in the form for a prolate spheroid (Ju and Sun, 1999, 2001; Mura,
1987; Sun and Ju, 2001). In case of an imperfect interface S, Qu (1993a) has derived a modified
Eshelby’s tensor SM ijkl , which accounts for the interfacial compliance. However, this modified
Eshelby’s tensor is no longer uniform but position dependent in the inhomogeneity
"0ij ðxÞ ¼ SM
ijkl ðxÞ"kl ð14Þ
which leads to a nonuniform perturbed strain field as well. Since the average perturbed strain over
the inhomogeneity is sufficient for this study, by taking the volume average of SM
ijkl ðxÞ over we may
arrive at
SM 0
ijkl ¼ Sijkl þ Sijmn Rmnpq Cpqst ðIstkl Sstkl Þ ð15Þ
where Istkl is the fourth-rank identity tensor (Lee and Pyo, 2009). Note that the derivation of SM
ijkl in
equation (15) does not involve an iterative calculation process which is necessary for higher order
solutions. Therefore, equation (15) is the simplified first-order solution and should be used in the
case of slightly weakened interface without excessive interfacial damage. The fourth-rank tensor
Rmnpq depends on the interfacial compliance and the geometry of the inhomogeneity, and can be
written as
Z
1
Rmnpq ¼ mp nq nn þ mq np nn þ np nq nm þ nq np nm dS ð16Þ
4 S
where "^0kl is the applied far-field uniform strain which is assumed identical to the average matrix
strain in the Mori–Tanaka approach. Through taking volume average on both sides of equation (17)
and rearranging the terms, we can reach
The average strain "1ij in the inhomogeneity, which is the summation of the far-field and average
perturbed strain, can thereby be obtained with the use of equation (13)
1
"1ij ¼ "0ij þ "0ij ¼ Iijkl þ SM 1
ijmn Amnkl "0kl : ð20Þ
where the contribution from interfacial displacement jump is also included. c0 and c1 are the volume
fractions of the inhomogeneity and the matrix, respectively. Since ij1 ¼ C1ijkl "1kl , combining equations
(20) and (21) we arrive at
h
i
"ij ¼ c1 Iijkl þ Rijmn C1mnkl þ c0 Iijkl þ SM 1
ijmn Amnkl "1kl ð22Þ
which is not affected by the interfacial condition. Finally the effective overall stiffness tensor C ijkl ,
which relates the overall average stress and strain as ij ¼ C ijkl "kl , can be calculated in terms of
equations (20), (22), and (23) as
1
C ijkl ¼ C0ijkl þ c1 C0ijmn A1
mnpq Rmnst C1
stpq c 1 Rpqrs C1
rskl þ I pqkl þ c0
M A1
Spqrs rskl ð24Þ
Note that the result in equation (24) gives the estimate for two-phase composites with aligned
ellipsoidal inhomogeneities where C ijkl is transversely isotropic. If all the inhomogeneities are ran-
domly oriented in the composites, an orientational averaging process is needed which is defined as
(Sun and Ju, 2004)
Z 2 Z =2
1
hi ðÞ sin d d
ð25Þ
2 0 0
After the above process the effective overall elastic stiffness tensor Cijkl of two-phase misoriented
composites becomes isotropic and can be expressed as
D E
D E1
C ijkl ¼ C0ijkl þ c1 C0ijmn A1 1
mnpq Rmnst Cstpq c1 Rpqrs C1rskl þ Ipqkl þ c0 SM 1
pqrs Arskl ð26Þ
If the correspondence principle is applied and the elastic tensors in equation (26) are replaced by
viscoelastic counterparts with complex components, the effective overall viscoelastic stiffness tensor
C ijkl is obtained.
parameters which have significant effect on these viscoelastic dynamic properties are investigated,
including the interfacial storage compliance 0 , the interfacial loss factor 00 =0 , and the aspect ratio
and concentration of MWNTs. The parameters for a typical type of elastomer and MWNTs used in
the simulations are listed in Table 1. Here the cylindrical MWNTs are regarded asisotropic elastic
ellipsoids with high aspect ratios, while the dynamic properties of the matrix elastomer are based on
one fixed frequency. It should be noted that in carbon nanotube reinforced composites, entangle-
ment and curvature of carbon nanotubes cannot be fully eliminated due to their high aspect ratio
and small size. Therefore, the above model may not apply to nanocomposites with highly non-
uniformly distributedcarbon nanotubes and high filler loadings.
ameter 0 is introduced to indirectly present 0 , following Qu (1993b) and Hashin (1990). This form
is especially useful in assigning values for 0 when frequency varies.
It can be observed that as 0 increases, the effective storage Young’s modulus E 0 of the nano-
composites drops monotonically regardless of 00 =0 . This is straightforward since increasing inter-
facial storage compliance leads to gradually weakened interfaces and therefore lower effective
stiffness. However, as 0 increases, different 00 =0 results in different trend of the effective loss
factor tan of the nanocomposites. If 00 =0 is lower than the matrix loss factor ðtan Þ0 , the com-
posite loss factor tan decreases; in case 00 =0 is higher than ðtan Þ0 , the composite tan decreases.
When perfect interfaces are assumed, the effective composite damping is not affected by MWNTs.
This phenomenon may be explained as follows. The increasing interfacial storage compliance grad-
ually initiates the contribution of interfacial damping in the effective composite damping, while
lower and higher interfacial damping will diminish and enhance the effective composite damping,
respectively.
In addition, it is also noticed in Figure 2 that the MWNT concentration affects the composite E 0
and damping. When MWNT concentration increases, the composite effective stiffness increases
correspondingly, while the trend of composite damping depends on interfacial damping.
Assuming weakened interfaces, when interfacial damping is lower than matrix damping, more
MWNTs lead to lower composite damping, and on the contrary, when interfacial damping is
higher than matrix damping, more MWNTs lead to higher composite damping.This might be due
to the more interfacial contact area resulted from more MWNTs.
Figure 2. The effect of interfacial storage compliance on: (a) composite storage Young’s modulus, (b) composite loss
factor (interfacial damping higher than matrix damping), and (c) composite loss factor (interfacial damping lower than
matrix damping). The parameters for MWNTs and interfaces are: MWNT radius a ¼ 20 nm, MWNT aspect
ratio ¼ 50, 0 ¼ 0 G0 =ð2aÞ, 00 =0 ¼ 0.2 or 00 =0 ¼ 0.05.
MWNTs: multi-walled carbon nanotubes.
Figure 3. The effect of interfacial loss factor on: (a) composite storage Young’s modulus, and (b) composite loss
factor. The
parameters for MWNTs and interfaces are: MWNT radius a ¼ 20 nm, MWNT aspect ratio ¼ 50,
0 ¼ 0 G0 =ð2aÞ ¼ 1.
MWNTs: multi-walled carbon nanotubes.
nanocomposites CNTs are very likely to be damaged, curved, and clustered, which makes it nearly
impossible to reach the high aspect ratios as they are synthesized. In this article, a range of 20–100
for MWNT aspect ratio is adopted in the simulation, and its effect is presented in Figure 4.
It is well known that increasing the particle aspect ratio of randomly oriented composites will
effectively improve the composite effective stiffness (e.g. Tandon and Weng, 1986). Here we can see
that for moderately weakened interfacial conditions, as the MWNT aspect ratio increases from 20 to
100, the composite E 0 is dramatically increased, which again reveals that CNTs are supreme stiffness
reinforcements for composites. Regarding the composite damping, increasing MWNT aspect ratio
will promote opposite trends given different interfacial damping, in that it leads to rising composite
damping if interfacial damping is lower than matrix damping, and dropping composite damping if
interfacial damping is higher than matrix damping.This is an interesting finding which predicts that
although CNTs with high aspect ratios are ideal reinforcements for composite stiffness, they may not
be that efficient in enhancing composite damping when the interfacial damping is relatively high.
Figure 4. The effect of MWNT aspect ratio on: (a) composite storage Young’s modulus, (b) composite loss factor
(interfacial damping higher than matrix damping), and (c) composite loss factor (interfacial damping
lower than matrix
damping). The parameters for MWNTs and interfaces are: MWNT radius a ¼ 20 nm, 0 ¼ 0 G0 =ð2aÞ ¼ 1,
00 =0 ¼ 0.2 or 00 =0 ¼ 0.05.
MWNTs: multi-walled carbon nanotubes.
The selection of interfacial parameters is a subtle and crucial issue. In the previous parametric
studies, the interfacial storage compliance 0 and loss factor 00 =0 are fixed at one frequency level to
rule out the frequency effect; however, in reality they might be both frequency dependent, reflecting
the viscoelastic nature of the interfaces under dynamic loads. In the available works studying the
creep and relaxation behavior of composites with viscoelastic interfaces, the interfacial parameters
have beenexpressedin the form of time-dependent relaxation compliance, which is determined by a
couple of arbitrarily selected prony terms (Gosz et al., 1990; Matzenmiller and Gerlach, 2004). In
this article, a different method is adopted to give interfacial parameters. As earlier mentioned, the
interfacial storage compliance 0 ð!Þ is associated withthe MWNT
radius a and the complex matrix
shear modulus G0 ð!Þ in the form of 0 ð!Þ ¼ 0 ð2aÞ=G0 ð!Þ. Since G0 ð!Þ is frequency dependent,
0 ð!Þ is naturally frequency dependent. As for the interfacial loss factor 00 ð!Þ=0 ð!Þ, it is assumed
constant over the entire frequency range to avoid complexity. Finally the interfacial complex com-
pliance is still frequency dependent.
Figure 5 shows the comparisons for effective composite E 0 and loss factor tan of nanocompo-
sites at 2 wt% and 5 wt% MWNT concentrations. The interfacial parameters selected are 0 ¼ 0.7
Figure 5. Comparison between simulation results and experimental data: (a) storage Young’s modulus; (b) loss
factor. The parameters
for MWNTs and interfaces in the simulation are: MWNT radius a ¼ 20 nm, MWNT aspect
ratio ¼ 40, 0 ¼ 0 G0 =ð2aÞ ¼ 0.7, 00 =0 ¼ 0.3.
MWNTs: multi-walled carbon nanotubes.
and 00 =0 ¼ 0.3. It is seen that the simulation results reasonably match the experimental data, from
which we may infer that the interfaces of the nanocomposites being investigated are somehow
weakened and of higher damping than the matrix elastomer.
Concluding remarks
A viscoelastic interface model has been proposed in this study, which allows for convenience in
examining the dynamic viscoelastic behavior of nanocomposites with imperfect interfaces between
filler CNTs and matrix polymer. Mori–Tanaka approach has been applied to derive the effective
dynamic stiffness and loss factor of randomly oriented nanocomposites, with the viscoelastic inter-
face model incorporated. Numerical simulations have been performed to investigate the effect on the
composite dynamic behavior due to a couple of factors including interfacial storage compliance,
interfacial damping, CNT aspect ratio, and CNT concentration, which all have turned out playing
important roles. Finally the proposed model has been verified by comparing simulation results and
experimental data.
Funding
This work was sponsored by the National Science Foundation under Grant No. CMMI-0800417. The support
of the NSF is gratefully acknowledged.
References
Aboudi J (1987) Damage in composites - modeling of imperfect bonding. Composites Science and Technology
28: 103–128.
Achenbach JD and Zhu H (1989) Effect of interfacial zone on mechanical-behavior and failure of fiber-
reinforced composites. Journal of the Mechanics and Physics of Solids 37: 381–393.
Achenbach JD and Zhu H (1990) Effect of interphases on micro and macromechanical behavior of hexagonal-
array fiber composites. Journal of Applied Mechanics-Transactions of the ASME 57: 956–963.
Kim BR and Lee HK (2011) Elastic-damage modeling for particulate composites considering cumulative
damage. International Journal of Damage Mechanics 20: 131–158.
Lee HK and Pyo SH (2008) Multi-level modeling of effective elastic behavior and progressive weakened
interface in particulate composite. Composites Science and Technology 68: 387–397.
Lee HK and Pyo SH (2009) 3D-damage model for fiber-reinforced brittle composites with microcracks and
imperfect interfaces. Journal of Engineering Mechanics 135: 1108–1118.
Li J and Weng GJ (1996) Effect of a viscoelastic interphase on the creep and stress/strain behavior of fiber-
reinforced polymer matrix composites. Composites Part B-Engineering 27: 589–598.
Li R and Sun LZ (2011) Dynamic mechanical analysis of silicone rubber reinforced with multi-walled carbon
nanotubes. Interaction and Multiscale Mechanics 4: 239–245.
Liu H and Brinson LC (2006) A hybrid numerical-analytical method for modeling the viscoelastic properties of
polymer nanocomposites. Journal of Applied Mechanics-Transactions of the ASME 73: 758–768.
Liu HT, Sun LZ and Ju JW (2004) An interfacial debonding model for particle-reinforced composites.
International Journal of Damage Mechanics 13: 163–185.
Liu HT, Sun LZ and Ju JW (2006) Elastoplastic modeling of progressive interfacial debonding for particle-
reinforced metal matrix composites. Acta Mechanica 181: 1–17.
Matzenmiller A and Gerlach S (2004) Micromechanical modeling of viscoelastic composites with compliant
fiber-matrix bonding. Computational Materials Science 29: 283–300.
Mori T and Tanaka K (1973) Average stress in matrix and average elastic energy of materials with misfitting
inclusions. Acta Metallurgica 21: 571–574.
Mura T (1987) Micromechanics of Defects in Solids. New York: Springer.
Pagano NJ and Tandon GP (1990) Modeling of imperfect bonding in fiber reinforced brittle matrix composites.
Mechanics of Materials 9: 49–64.
Paulino GH, Yin HM and Sun LZ (2006) Micromechanics-based interfacial debonding model for damage of
functionally graded materials with particle interactions. International Journal of Damage Mechanics 15:
267–288.
Qu JM (1993a) Eshelby tensor for an elastic inclusion with slightly weakened interface. Journal of Applied
Mechanics-Transactions of the ASME 60: 1048–1050.
Qu JM (1993b) The effect of slightly weakened interfaces on the overall elastic properties of composite-materi-
als. Mechanics of Materials 14: 269–281.
Schandler LS, Brinson LC and Sawyer WG (2007) Polymer nanocomposites: a small part of the story. JOM 59:
53–60.
Sun LZ and Ju JW (2001) Effective elastoplastic behavior of metal matrix composites containing randomly
located aligned spheroidal inhomogeneities, Part II: Applications. International Journal of Solids and
Structures 38: 203–225.
Sun LZ and Ju JW (2004) Elastoplastic modeling of metal matrix composites containing randomly located and
oriented spheroidal particles. ASME Journal of Applied Mechanics 71: 774–785.
Sun LZ, Ju JW and Liu HT (2003a) Elastoplastic modeling of metal matrix composites with evolutionary
particle debonding. Mechanics of Materials 35: 559–569.
Sun LZ, Liu HT and Ju JW (2003b) Effect of particle cracking on elastoplastic behavior of metal matrix
composites. International Journal for Numerical Methods in Engineering 56: 2183–2198.
Sun L, Gibson RF, Gordaninejad F, et al (2009) Energy absorption capability of nanocomposites: a review.
Composites Science and Technology 69: 2392–2409.
Tan H, Huang Y and Liu C (2008) The viscoelastic composite with interface debonding. Composites Science
and Technology 68: 3145–3149.
Tandon GP and Weng GJ (1986) Average stress in the matrix and effective moduli of randomly oriented
composites. Composites Science and Technology 27: 111–132.
Wang X, Li Q, Xie J, et al (2009) Fabrication of ultralong and electrically uniform single-walled carbon
nanotubes on clean substrates. Nano Letters 9: 3137–3141.
Xie XL, Mai YW and Zhou XP (2005) Dispersion and alignment of carbon nanotubes in polymer matrix: a
review. Materials Science & Engineering R-Reports 49: 89–112.
Appendix
The fourth-rank tensor Rijkl
By substituting equation (3) into equation (16), the fourth-rank tensor Rijkl can be written as
where for ellipsoidal inclusions with the symmetry axis identified as axis 1, the components of Pijkl
and Qijkl are
Z Z 2
3 d
Pijkl ¼ ik n^ j n^ l þ jk n^ i n^ l þ il n^ j n^ k þ jl n^ i n^ k sin d ð28Þ
16 0 0 n
Z Z 2
3 n^i n^j n^k n^l
Qijkl ¼ d sin d ð29Þ
4 0 0 n
pffiffiffiffiffiffiffiffi sin cos sin sin cos T
with n ¼ n^ i n^ i and n^i ¼ , ,
a1 a2 a3